Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Advances in Materials Research, Vol. 1, No. 2 (2012) 129-146 129
Modeling and multiple performance optimization ofultrasonic micro-hole machining of PCD usingfuzzy logic and taguchi quality loss function
Vinod Kumar*1 and Neelam kumari
2
1Department of Mechanical Engineering, Thapar University Patiala, PUNJAB, 147004, India2Department of Statistics, Punjabi University Patiala, PUNJAB, 147002, India
(Received April 10, 2012, Revised June 15, 2012, Accepted June 21, 2012)
Abstract. Polycrystalline diamond is an ideal material for parts with micro-holes and has been widelyused as dies and cutting tools in automotive, aerospace and woodworking industries due to its superiorwear and corrosion resistance. In this research paper, the modeling and simultaneous optimization ofmultiple performance characteristics such as material removal rate and surface roughness of polycrystallinediamond (PCD) with ultrasonic machining process has been presented. The fuzzy logic and taguchi’squality loss function has been used. In recent years, fuzzy logic has been used in manufacturingengineering for modeling and monitoring. Also the effect of controllable machining parameters like typeof abrasive slurry, their size and concentration, nature of tool material and the power rating of themachine has been determined by applying the single objective and multi-objective optimizationtechniques. The analysis of results has been done using the MATLAB 7.5 software and results obtainedare validated by conducting the confirmation experiments. The results show the considerable improvementin S/N ratio as compared to initial cutting conditions. The surface roughness of machined surface has beenmeasured by using the Perthometer (M4Pi, Mahr Germany).
Keywords: fuzzy logic; micro-machining; surface roughness; modeling; optimization; ultrasonic
1. Introduction
Modern materials such as high-strength metals and ceramics that are developed to meet the needs
of advanced industries are typically strong, hard and brittle. There has been the introduction of
many new materials such as tungsten and titanium carbides, polycrystalline diamonds, rubies,
sapphire, hard steels, magnetic alloys and corundum. Polycrystalline diamond is having high
thermal conductivity, high wear resistance, high hardness, high electrical conductivity and high
resistance to corrosion. The material removal rate and surface roughness are important parameters in
ultrasonic machining process. While technologically desirable, these characteristics often render the
materials difficult and sometimes impossible to shape by machining processes into useful
components and parts.
The use of hard and brittle materials has become increasingly more extensive. However, it is not
*Corresponding author, Ph. D., E-mail: [email protected]
DOI: http://dx.doi.org/10.12989/amr.2012.1.2.129
130 Vinod Kumar and Neelam kumari
feasible to machine these materials with the application of traditional metal-cutting techniques. The
processing of such materials for the part fabrication has become a challenging problem. However, if
ultrasonic energy is applied to the machining process and coupled with the use of hard abrasive grits,
extremely hard and brittle materials can be effectively machined. The methods of optimization can be
classified into two approaches namely reliability based and robust design based methods. The
objective of robust design is to optimize the mean and minimize the variability that results from
uncertainty represented by noise factors. The robust design of a vibration absorber with mass and
stiffness uncertainty in the main system is used to demonstrate the robust design approach in
dynamics as reported by Kumar and Khamba (2006). The circularity, cylindricity, surface roughness
and hole oversize of the ultrasonically and conventionally drilling of Inconel 738-LC were measured
and compared by Azarhoushang and Akbari (2007). The on-line tool wear monitoring during
ultrasonic machining using tool resonance frequency was determined by Hocheng and Kuo (2002).
The effects of various parameters of ultrasonic drilling of two-dimensional carbon fiber-reinforced
silicon carbide(C/SiC) composites including abrasives, volume ratio, electric current and down-force
on the material removal rate, hole clearance, edge quality and tool wear were studied by Hocheng et
al. (2000). The optimum parameters for multi-performance characteristics in drilling using grey
relational analysis were determined by Tosun (2006). The design optimization of cutting parameters
for side milling operations with multiple performance characteristics was done by Chang and Lu
(2007). The experiments were conducted to understand the tool wear mechanism in rotary ultrasonic
machining (RUM) of silicon carbide (SiC). The topography of the end face and lateral face of a
diamond tool in RUM of SiC was observed under digital microscope by Zeng et al. (2005). The laser
processing of polycrystalline diamond, tungsten carbide and a related composite material was done
by Harrison and Henry (2006). The parametric optimization of ultrasonic machining of Co-based
super alloy using the taguchi multi-objective approach has been done by Kumar and Khamba (2009).
The robust design method is essential for improving engineering productivity as reported by Roy
(1990). The statistical analysis of experimental parameters in ultrasonic machining of tungsten
carbide using taguchi approach has been done by Kumar and Khamba (2008). The grey taguchi
method was applied to optimize the milling parameters by Tsao (2009). The effect of various
machining parameters during machining of PCD was studied by Tso and Lin (2002). From the
literature review it has been concluded that the modeling and multi-objective optimization of
parameters involved in ultrasonic micro machining of PCD has not been done by the previous
authors. Therefore as the polycrystalline diamond is an ideal material for parts with micro-holes and
has been widely used as dies and cutting tools in automotive, aerospace and woodworking industries
due to its superior wear and corrosion resistance. In view of the extensive applications, there was the
need of this type of research work to be done. In the present study, the multiple performance
characteristics have been optimized simultaneously and more informed analysis has been made by
conducting less expensive experiments for design improvement. Also the modeling and simultaneous
optimization of the results has been done using the fuzzy logic and taguchi quality loss function.
2. Taguchi analytical methodology
Taguchi’s method of experimental design provides a simple, efficient and systematic approach to
determine optimal machining parameters as studied by Gaitonde et al. (2007). Taguchi has
recommended orthogonal arrays (OA) for the designing of experiments. In taguchi method, the
Modeling and multiple performance optimization of ultrasonic micro-hole machining 131
results of experiments are analyzed to achieve one or more of the objectives as to establish the best
or the optimum condition for a product or process, to estimate the contribution of individual
parameters and interactions and to estimate the response under the optimum condition. The
optimum condition for hole roundness in deep holes has been found by Deng and Chin (2005).
Analysis of variance (ANOVA) is the statistical treatment applied to the results of the experiments in
determining the percent contribution of each parameter against a stated level of confidence. The study
of ANOVA table for a given analysis helps to determine which of the parameters need control and
which do not. Taguchi suggested two different routes to carry out the complete analysis. First, the
standard approach; where the results of a single run or the average of repetitive runs are processed
through main effect and analysis of variance. The second approach, which taguchi strongly
recommends for multiple runs, is to use signal-to-noise ratio (S/N) for the same steps in the analysis.
The S/N ratio is a concurrent quality metric linked to the loss function as reported by Phadke (1989).
Design of experiment (DOE) methods result in an efficient experimental schedule and produce
a statistical analysis to determine easily as to which parameters have the most significant effects
on the final results. The use of signal-to-noise ratio (S/N) in system analysis provides a
quantitative value for response variation comparison. The requirement to test multiple factors
means that a full factorial experimental design that describes all possible conditions would result
in a large number of experiments. After conducting the experiments, the data from all
experiments has to be evaluated to determine the optimum levels of the design variables using
the analysis of variance (ANOVA) and the analysis of mean (ANOM) of the S/N ratio. There are
several S/N ratios available depending on the types of characteristics; lower is better (LB),
nominal is best (NB) and higher is better (HB).
Lower-the-better type problem
(1)
where (S/N)L is signal-to-noise ratio for lower-the-better type problem, n is the number of repetitions
for each trial, independent of the values assigned to noise factors, and yi is the value of the response
obtained in the ith repetition of the trial.
Higher-the-better type problem
In this type of problem, the quality characteristic is again continuous and non-negative and it is to
be made as large as possible. There is no adjustment factor to be used in this case as well and one
is interested in maximizing the objective function expressed as
(2)
Where (S/N)H is signal-to-noise ratio for higher-the-better type problem.
Nominal-the-best type problem
In the nominal-the-best type problem, the quality characteristic is continuous and non-negative,
but its target value is non zero and assumes some finite value. For these types of problems, if the
mean becomes zero the variance also tends to become zero. A scaling factor can be used as an
adjustment factor to shift the mean closer to the target for such type of problems. The objective
function that is to be maximized can be expressed as
S N⁄( )L 10 1 n yi
2
i 1=
n
∑⁄⎝ ⎠⎜ ⎟⎛ ⎞
10log–=
S N⁄( )H 101
n---
1
yi
2----
i 1=
n
∑⋅⎩ ⎭⎨ ⎬⎧ ⎫
10
log–=
132 Vinod Kumar and Neelam kumari
(3)
Where (S/N)N is signal-to-noise ratio for Nominal-the-best type problem
(4)
(5)
3. Tool design for experimentation
In ultrasonic machining, the mass and dimensions of the tool constitute a very important design
problem to economize the machining operation. As the tool materials selected for the experimentation
possess different densities, the designing of the tools is needed to be done with a consideration that
the mass of each tool should be same to the maximum possible extent. From the pilot
experimentation, it was concluded that mass of the tool should be within the critical limit of 50 gm.
All the tools were made as single piece unit by machining on a centre lathe. The tip of tool contains
unified threads and is tightened to the horn manually. The horn is of 25.4 mm size and it contains
internal threads. The length of the tool tip in ultrasonic machining process needs to be restricted and
maintained within the limits of 15-20 mm for optimum results. The use of a length more than 20
mm resulted in over stressing of the tool and shortened tool life.
While designing the tools, the dimensions for each tool were decided to ensure that the mass of
each tool is same and is equal to 50 gm with a permissible variation of one gm. The tool finish is
another important factor that is found to affect the surface finish of the machined surface. Hence,
the surface finish of tool face was maintained at a level of 4.5 microns before starting a new
experiment. The tool face also tends to gain a convex shape as a result of uneven distribution of the
abrasive particles under the tool face while machining takes place. This alters the contour of the
machined surface as well as the material removal rate. To rectify this problem, facing operation was
performed on each tool on center lathe after a particular experimental run was executed. This helped
to ensure a perfectly flat surface of the tool which is responsible for machining and thus the
undesirable effect on the shape and size of the cavity produced is also controlled. Further, in order
to deal with the problems of fatigue failure of the tools while machining and other problems
pertaining to the overheating and stress loading of the tools, a number of tools were prepared for
each tool material. This also helped to maintain the continuity of the experimentation.
4. Experimentation
The experiments were performed on a Sonic-Mill, 500 W (Albuquerque, NM) as shown in Fig. 1.
The machining of work material was performed using different input parameters the tool material
being three different titanium alloys (TITAN12, TITAN15 and TITAN31). The frequency was
varied from 18 to 22 kHz. The three different values of power rating taken were (25, 50 and 75)
S N⁄( )N 10 µ2 σ2⁄( )10log=
µ1
n--- yi
2
i 1=
n
∑⋅=
σ1
n 1–( ) yi µ–( )2
i 1=
n
∑⋅
---------------------------------------------=
Modeling and multiple performance optimization of ultrasonic micro-hole machining 133
percent. The three different abrasive slurries (Al2O3, SiC and B4C), each of three grit sizes (220,
320 and 500) were adopted with percentage concentrations by volume with water (20, 25 and 30).
The Table 1 shows the control variables and their levels. There was no withdrawal of the tool
during the tests. Abrasive slurry feed circulation and frequency amplitude was maintained
constant. The frequency measurement was performed with the help of a frequency meter. The
trials were carried out under maximum material removal rate (MRR) conditions with a tool
Fig. 1 Pictorial view of the experimental set-up
Table 1 Representation of control variables and their levels
S.No Control variables Levels Level 1 Level 2 Level 3
A Tool material 3 TITAN12 TITAN15 TITAN31
B Abrasive slurry 3 Al2O3 SiC B4C
C Slurry concentration (%) 3 20 25 30
D Abrasive grit size 3 220 320 500
E Power rating (%) 3 25 50 75
134 Vinod Kumar and Neelam kumari
rotation of 350 rpm. All the experiments were repeated four times; hence four trials were
conducted at each experimental run. The output variables were recorded for each trial and then
the results for each experimental run were averaged out to obtain the mean value of response
variable (MRR) for that particular experiment. The analysis of results has been performed using
the MATLAB 7.5 software.
5. Analysis and verification of results
5.1 Selection of orthogonal arrays
The orthogonal array based on the taguchi concept was utilized to arrange the discrete variables
and robust solutions for unconstrained optimization problems were found. In this investigation, the
five machining parameters, tool material, abrasive slurry, slurry concentration, grit size and power
rating were taken with three different levels of each. Thus a total of 243 (3×3×3×3×3) different
combinations were considered. However, according to taguchi, the samples could be organized into
only 18 groups and if they were to be considered separately still it yield results with the same
Table 2 S/N ratios for MRR and SR in single quality optimization
Exp.No
Average MRR(µm3/min)
AverageSR(µm)
S/N ratio(dB)
MRR SR
1 0.186 0.795 14.609 1.992
2 0.200 1.075 13.979 0.628
3 0.218 0.995 13.230 0.043
4 0.229 0.755 12.803 2.441
5 0.229 1.01 12.803 0.086
6 0.218 0.995 13.230 0.043
7 0.245 0.49 12.216 6.196
8 0.267 1.175 11.469 1.400
9 0.238 1.185 12.468 1.185
10 0.232 1.16 12.690 1.474
11 0.229 0.895 12.803 0.895
12 0.254 1.627 11.903 0.963
13 0.259 1.175 11.734 1.400
14 0.254 0.825 11.903 1.670
15 0.250 1.085 12.041 0.708
16 0.255 1.285 11.869 2.178
17 0.123 0.555 18.201 5.114
18 0.132 0.555 17.588 5.114
Modeling and multiple performance optimization of ultrasonic micro-hole machining 135
confidence. The S/N ratios of MRR and surface roughness in single quality optimization according
to the arrangement of the samples into 18 groups; L18 according to taguchi is shown in Table 2. The
numbers (1, 2 and 3) represents the various experimental levels of the different factors. The initial
parameter settings of the experiment “A1B1C1D1E1” was decided from the pilot experimentation
done to determine the significant parameters.
5.2 Determination of quality loss for each quality characteristics
The material removal rate is larger-the-better type and surface roughness is the smaller-the better
type. A quality loss or mean square deviation (MSD) function is used to calculate the deviation
between the experimental value and the desired value. The mean square deviation is different for
the different types of problems.
Smaller-the better type
(6)MSDy1
2y2
2y3
2y4
2 … yn
2+ + + + +
n---------------------------------------------------------=
Table 3 Computational results of the parameters
Quality loss Normalized quality loss Total normalized quality loss and multiple S/N ratio
MRR SR MRR SR TNQL MSNR(dB)
28.905 0.632 0.437 0.238 0.357 4.473
25.000 1.155 0.378 0.436 0.401 3.968
21.042 0.990 0.318 0.374 0.340 4.685
19.069 0.570 0.288 0.215 0.258 5.883
19.069 1.020 0.288 0.385 0.326 4.867
21.042 0.990 0.318 0.374 0.340 4.685
16.659 0.240 0.252 0.090 0.187 7.281
14.027 1.380 0.212 0.521 0.335 4.749
17.654 1.404 0.267 0.530 0.372 4.294
18.579 1.345 0.281 0.508 0.371 4.306
19.069 0.801 0.288 0.302 0.293 5.331
15.50 2.647 0.234 1.000 0.540 2.676
14.907 1.380 0.225 0.521 0.343 4.647
15.50 0.680 0.234 0.256 0.242 6.161
16.0 1.177 0.242 0.444 0.322 4.921
15.378 1.651 0.232 0.623 0.388 4.111
66.098 0.308 1.000 0.116 0.646 1.938
57.392 0.308 0.868 0.116 0.567 2.464
Mean MSNR(nm) 4.524
136 Vinod Kumar and Neelam kumari
Higher-the better type
(7)
Where represents the responses of experiments, n is the number of repetitions.
The quality loss values for each quality characteristics are shown in Table 3.
5.3 Determination of normalized quality loss for each quality characteristic
The normalized quality loss has been determined by using following formula
(8)
Where represents the normalized quality loss, is quality loss for the quality
characteristic at the run in the experiment design matrix and is maximum quality loss for the
quality characteristic among all the experimental runs. The normalized quality loss values for
each quality characteristics are shown in Table 3.
5.4 Determination of total normalized quality loss
The total normalized quality loss has been determined by using following formula
(9)
Where represents the total normalized quality loss, is normalized quality loss for the
quality characteristic at the run in the experiment design matrix, is the weighting factor for
the quality characteristic and k is number of quality characteristics. Here k=2 and assuming
weighting factors z for MRR and SR as 0.6 and 0.4. The total normalized quality loss (TNQL) and
multiple S/N ratio (MSNR) are shown in Table 3.
5.5 Determination of multiple S/N ratio, factor effects and optimum combinations
The multiple S/N ratios as given in Table 4 have been determined from the following formula
(10)
The optimum combinations corresponding to maximum average effect are considered. The
optimum combination of parameters is A1B3C1D3E3.
5.6 Performing the confirmatory experiments
The confirmatory experiments has been performed with optimum settings of the factors and levels
as determined to verify the optimum conditions. The multiple S/N ratio at optimum level has been
determined by applying the following formula
MSD1 y1
21 y2
21 y3
21 y4
2 … 1 yn
2⁄+ +⁄+⁄+⁄+⁄n
---------------------------------------------------------------------------------------=
y1 y2 y3 y4 and yn, , ,
yij
Lij
Lim
-------=
yij Lij ith
jth
Lim
ith
Yj ziyij
i 1=
k
∑=
Yj yij ith
jth
ziith
ηj 10 Yj( )10
log–=
Modeling and multiple performance optimization of ultrasonic micro-hole machining 137
(11)
Where represents the average value of multiple S/N ratios in all experimental runs, are
multiple S/N ratios corresponding to optimum factor levels and p is the number of factors.
The predicted multiple S/N ratio and that from the confirmatory experiments is shown in Table 5.
The improvement in multiple S/N ratio at the optimum combination is found to be 1.481 dB. The
values of material removal rate and surface roughness at this optimum combination are 0.201 (µm3/
min) and 0.522 (µm) in comparison to 0.186 (µm3/min) and 0.795 (µm) for initial setting of
parameters.
5.7 Comparison of multi-objective and single objective optimization results
The results of single quality optimization for MRR and surface roughness are summarized in
Tables 6 and 7. The confirmatory experiments results of single objective optimization are shown in
Table 8. The results of multi-objective optimization (MOO) and single objective optimization (SOO)
using taguchi quality function has been compared in Table 9. The results shows that the quality
values at optimum settings are different in each case. The results of MOO basically depend on
weights assigned to quality values. As in the present research work, the most important quality
assumed was MRR with weight 0.6 and the optimum MRR value in SOO is more as compared to
MRR obtained from MOO. The result is almost same to that of optimum SR (obtained SOO) while.
Therefore chance of quality loss is always there, when the objective is to optimize the multiple
ηop ηm ηi ηm–( )i 1=
p
∑+=
ηm ηi
Table 4 Effect of factor levels on multiple S/N ratio
FactorsMean MSNR(dB)
Level 1 Level 2 Level 3
Tool material 5.116* 4.502 3.954
Abrasive slurry 4.473 4.317 4.782*
Slurry concentration 5.065* 4.384 4.122
Abrasive grit size 3.861 4.512 5.197*
Power raring 4.430 4.516 4.626*
*Optimum level
Table 5 Confirmatory experiments results (Multi-objective optimization)
Predicted Experimental
Level A1B1C1D1E1 A1B3C1D3E3 A1B3C1D3E3
MRR 0.186 - 0.201
SR 0.795 - 0.522
MSNR(dB) 4.473 4.874 5.954
Improvement of MSNR: 1.481 dB
138 Vinod Kumar and Neelam kumari
Table 6 S/N response table for MRR in single quality optimization
FactorsMean S/N ratio(dB)
Level 1 Level 2 Level 3
Tool material 12.698 13.376* 11.426
Abrasive slurry 13.820* 13.370 12.398
Slurry concentration 13.724* 12.344 13.521
Abrasive grit size 13.524 13.571* 12.493
Power rating 12.592 13.726* 13.271
*Optimum level
Table 7 S/N response table for SR in single quality optimization
FactorsMean S/N ratio(dB)
Level 1 Level 2 Level 3
Tool material 2.613* 1.632 1.342
Abrasive slurry 1.906 2.397* 1.624
Slurry concentration 2.651* 1.059 1.550
Abrasive grit size 1.818 2.085* 1.684
Power rating 1.345 2.360* 1.882
*Optimum level
Table 8 Confirmatory experiments results (single objective optimization)
Predicted Experimental
Level A1B1C1D1E1 A1B3C1D3E3 A1B3C1D3E3
MRR 0.186 - 0.224
SR 0.795 - 0.498
S/N(dB) for MRR 14.609 16.242 16.044
S/N(dB) for SR 1.992 2.142 2.032
Improvement of S/N for MRR: 1.435 dB Improvement of S/N for SR: 0.04 dB
Table 9 Comparison of results from single and multi-objective optimization
SOO results MOO resultsQuality loss (%)
MRR SR MRR & SR
Level A2B1C1D2E2 A1B2C1D2E2 A1B3C1D3E3
MRR 0.224 - 0.201 11.443
SR - 0.498 0.522 4.597
Modeling and multiple performance optimization of ultrasonic micro-hole machining 139
quality characteristics simultaneously. The multi-objective optimization is useful in the sense that at
the same optimum parameter level, one can get the optimum quality value of multiple quality
characteristics at the same time rather than a single optimum quality characteristic. The ANOVA for
MRR and SR is shown in Tables 10 and 11.
6. Modeling of results using fuzzy logic approach
Fuzzy logic is an effective tool for dealing with complex nonlinear systems. Fuzzy logic is based on
imprecision and is similar to the way people make decisions based on imprecise and non numerical
information. Fuzzy logic modeling is based on mathematical theory combining multivalued logic,
probability theory and artificial intelligence methods. Fuzzy modeling is based on fuzzy set theory in
which the linguistic statements are expressed mathematically and corresponds to the analysis of a
human expert. The inputs and outputs in fuzzy systems are in the form of linguistic variables. The
variables are then tested with IF-THEN rules, which produce one or more responses depending on
Table 10 Analysis of variance (MRR)
Effect SS F P-value
Tool material 5.17 0.70 0.549
Abrasive slurry 19.32 2.62 0.188
Slurry concentration 1.78 0.24 0.796
Abrasive grit size 5.49 0.74 0.531
Power rating 4.03 0.55 0.617
Tool material X Abrasive slurry 29.67 2.01 0.258
Tool material X SC 7.79 0.53 0.72
Abrasive slurry X SC 3.77 0.26 0.892
Residual error 14.76
Table 11 Analysis of variance (SR)
Effect SS F P-value
Tool material 6.50 0.62 0.584
Abrasive slurry 15.52 1.47 0.332
Slurry concentration 2.31 0.22 0.813
Abrasive grit size 6.50 0.62 0.585
Power rating 19.06 1.81 0.276
Tool material X Abrasive slurry 16.84 0.80 0.584
Tool material X SC 63.31 3.00 0.156
Abrasive slurry X SC 2.58 0.12 0.967
Residual error 21.11
140 Vinod Kumar and Neelam kumari
which rules are asserted. Each rule has an antecedent part and a consequent part. The antecedent part is
a collection of conditions connected by AND, OR, NOT logic operators and consequent part represents
its action. In fuzzy inference engine, the truth value for the premise of each rule is computed and
applied to conclusion part of each rule. This results in one fuzzy subset being assigned to each output
variable for each rule. The response of each rule is weighed according to the degree of membership of
its inputs and the centroid of the responses is calculated to generate the appropriate output.
The concept of fuzzy reasoning for three input one output fuzzy logic unit is described as follows.
The fuzzy rule base consists of a group of IF-THEN statements with three inputs x1,x2,x3 and one
output y; that is,
Rule 1: if x1 is A1 and x2 is B1 and x3 is C1 then y is D1; else
Rule 2: if x1 is A2 and x2 is B2 and x3 is C2 then y is D2; else
Rule 3: if x1 is A3 and x2 is B3 and x3 is C3 then y is D3; else
Rule n: if x1 is An and x2 is Bn and x3 is Cn the y is Dn;
Ai, Bi, Ci and Di are fuzzy subsets defined by corresponding membership functions; that is µA, µB,
µC and µD.
Eighteen rules were developed based on experimental conditions. By taking max-min
compositional operation, the fuzzy reasoning of these rules yields a fuzzy output. Suppose that x1,x2
and x3 are the three input variables of the fuzzy logic unit, the membership function of the output of
fuzzy reasoning can be expressed as
(12)
Where is the minimum operator and V is the maximum operator.
The membership functions can be of different forms like triangular, trapezoidal, Gaussian, sigmoid
etc. In this study, triangular and trapezoidal membership functions are considered. The triangular
shaped membership function for input is specified by three parameters {a,b,c} as follows
(13)
By using min and max, an alternate expression for the proceeding equation is
(14)
Where a, b, c stand for the triangular fuzzy triplet and determine the x coordinates of the three
corners of the underlying triangular membership function.
The input-output numerical values are correlated by linguistic variables. This was obtained
through the design of membership functions consisting of fuzzy set values. The linguistic values
such as LOW, MEDIUM and HIGH are used to represent the input variables slurry concentration,
grit size and power rating. The output numerical values are also correlated in a similar manner, by
means of membership functions such as LOWEST, LOWER, LOW, LOW MEDIUM, MEDIUM,
HIGH MEDIUM, HIGHER, HIGH and HIGHEST. The membership functions used in this work
µD0
y( ) µA1
x1( )ΛµB1
x2( )ΛµC1
x3( )ΛµD1
y( )V… µAn
x1( )ΛµBn
x2( )ΛµCn
x3( )ΛµDn
y( )×[ ]=
Λ
triangle x a b c, ,;( )
0 x a≤,
x a–
b a–----------- a x b≥≤
c x–
c b–---------- b x c≥≤
0 c x≤,⎩⎪⎪⎪⎨⎪⎪⎪⎧
=
triangle x a b c, ,;( ) max minx a–
b a–-----------
c x–
c b–----------,⎝ ⎠
⎛ ⎞ 0,⎝ ⎠⎛ ⎞=
Modeling and multiple performance optimization of ultrasonic micro-hole machining 141
using a triangular membership function for input parameters slurry concentration, grit size and
power rating and the output parameters material removal rate (MRR) and surface roughness are
represented in Fig. 2. From the figure, one can infer that the experimental values and fuzzy values
Fig. 2 Membership functions for input and output parameters using triangular membership function (slurryconcentration, grit size and power rating)
142 Vinod Kumar and Neelam kumari
are very close to each other and hence the fuzzy rule based modeling technique can be effectively
used for prediction of MRR and surface roughness.
Similarly, the trapezoidal shaped membership function for input is specified by four parameters as
follows
Fig. 3 Membership functions for input and output parameters using trapezoidal membership function (slurryconcentration, grit size and power rating)
Modeling and multiple performance optimization of ultrasonic micro-hole machining 143
(15)
An alternative expression using min and max is
(16)
The membership functions used in this work using a triangular membership function for input
parameters slurry concentration, grit size and power rating and the output parameters material
removal rate (MRR) and surface roughness are represented in Fig. 3. He parameters [a, b, c, d]
determine the x coordinates of the four corners of the underlying trapezoidal membership function.
Finally, a defuzzification method is used. Defuzzification is an important operation in the theory of
fuzzy sets. It transforms fuzzy set information into numeric information. In the present study, the
centroid defuzzification method has been selected, because it produces the center of area of
possibility distribution of the inference output and is a more frequently used defuzzification method
for calculating the centroid of the area under the membership function
(17)
The non fuzzy value y0 gives the output value in numerical form. The comparison between the
trapezoid x a b c d, , ,;( )
0 x a≤,
x a–
b a–----------- a x b≤ ≤
1 b x c≤ ≤,
d x–
d c–---------- c x d≥≤
0 d x≤,⎩⎪⎪⎪⎪⎨⎪⎪⎪⎪⎧
=
trapezoid x a b c d, , ,;( ) max minx a–
b a–----------- 1
d x–
d c–----------, ,⎝ ⎠
⎛ ⎞ 0,⎝ ⎠⎛ ⎞=
y0
yµDo
y( )∑µD
o
y( )∑------------------------=
Fig. 4 Comparison between experimental results and fuzzy results for MRR (µm3/min)
144 Vinod Kumar and Neelam kumari
experimental and the fuzzy model prediction values for material removal rate and surface roughness
is presented in Figs. 4 and 5. From the figure, one can infer that the experimental values and fuzzy
values are very close to each other and hence the fuzzy rule based modeling technique can be
effectively used for prediction of MRR and surface roughness in ultrasonic drilling of Titanium
alloys.
7. Results and discussion
The material removal rate is of primary importance in rough ultrasonic machining of
polycrystalline diamond. This study confirms that there exists an optimum condition for precision
machining of PCD although the condition may vary with the composition of the material, the
accuracy of the machine and other external factors. The taguchi quality function has been applied
because high material removal rate and low surface roughness are conflicting goals, which cannot
be achieved simultaneously with a particular combination of control settings.
It was observed that for the PCD, MRR tend to increase with a corresponding increase in the
coarseness of the slurry used irrespective of the abrasive used for preparation of the slurry. This is
because the coarser grit causes more extensive damage to the material during the abrasive impact.
When the size of the abrasive particle becomes comparable with the tool amplitude, maximum
MRR is obtained. Any further increase in grit number decreases the grit size considerably, resulting
in several layers of abrasive particles which results in less effective machining. Also, the MRR
obtained with different tool materials (TITAN12, TITAN15 and TITAN31) are significantly different
when all other input parameters are controlled and remain fixed. Thus, the tool material properties
such as hardness and toughness also have been found to control the machining characteristics in
USM of PCD.
The optimum combination of design parameters is A1B3C1D3E3 as shown in Table 4. The test
results reveal the following as optimum operating conditions: a tool material of TITAN12, a slurry
concentration of 20%, a grit size of 500, abrasive slurry of B4C and a power rating of 75% (375 W).
A mathematical model of the material removal rate using fuzzy logic appraoch has been
Fig. 5 Comparison between experimental results and fuzzy results for surface roughness (µm)
Modeling and multiple performance optimization of ultrasonic micro-hole machining 145
formulated by identifying the physical parameters that affect the process of material removal and
surface roughness in ultrasonic machining process. The calculated results from the model show
good agreement when compared to the experimental findings.
8. Conclusions
1. The fuzzy logic rule based models for material removal rate and surface roughness were
developed for the experimental data using two different membership functions, viz. triangular and
trapezoidal. The predicted fuzzy output values and measured values are fairly close to each other,
which indicate that the fuzzy logic model can be effectively applied to predict the material removal
rate and surface roughness in ultrasonic machining of polycrystalline diamond.
2. In fuzzy rule based modeling, the trapezoidal membership functions perform better than
triangular membership functions.
3. The taguchi quality loss function can be used to optimize the multiple quality characteristics. The
quality characteristics experimental values of material removal rate and surface roughness at
optimum conditions (0.201 µm3/min, 0.522 µm) have been improved considerably in comparison to
initial parameter settings of the experiment (0.186 µm3/min, 0.795 µm). The improvement in MSNR
at the optimum combination found to be 1.481 dB.
4. The optimum parameter values in the present operating conditions found to be are tool material:
Titan 12, abrasive slurry: B4C, slurry concentration: 20%, abrasive grit size: 500 and power rating:
375 W.
5. The material removal rate and surface roughness have been affected by using the different types
of abrasive slurries. It could be concluded that use of boron carbide slurry results in better material
removal rate for same process conditions in comparison to silicon carbide and aluminum oxide.
This can be attributed to the higher hardness and cutting ability of boron carbide in comparison to
silicon carbide and aluminum oxide abrasives.
6. The loss of quality is always possible during optimization of multiple quality characteristics at a
time. The deviation of quality from its optimum value depends mainly on the weight assigned to it.
Therefore a careful selection of weights for different quality values plays a crucial role in multi-
objective optimization.
Acknowledgements
The authors would like to thank Dr. V.K. Jain (Professor, Mechanical Engineering department,
Indian institute of technology, Kanpur) and Mr. Charlie White (Sonic Mill, Albuquerque, USA) for
providing technical guidance for the experimentation work. Also the authors gratefully acknowledge
the authorities of Thapar University, Patiala for providing the Laboratory facilities utilized for the
experimentation work.
References
Azarhoushang, B. and Akbari, J. (2007), “Ultrasonic-assisted drilling of inconel 738-LC”, Int. J. Mach. Tool
146 Vinod Kumar and Neelam kumari
Manuf., 47(7-8), 1027-1033.Chang, C.K. and Lu, H.S. (2007), “Design optimization of cutting parameters for side milling operations with
multiple performance characteristics”, Int. J. Adv. Manuf. Tech., 32(1-2), 18-26.Deng, C.S. and Chin, J.H. (2005), “Hole roundness in deep hole drilling as analysed by taguchi methods”, Int. J.
Adv. Manuf. Tech., 25(5-6), 420-426.Gaitonde, V.N., Karnik, S. and Achyutha, B.T. (2007), “Methodology of taguchi optimization for multi-objective
drilling problem to minimize the burr size”, Int. J. Adv. Manuf. Tech., 34(1-2), 1-8.Hocheng, H. and Kuo, K.L. (2002), “On-line tool wear monitoring during ultrasonic machining using tool
resonance frequency”, J. Mater. Process. Tech., 123(1), 80-84.Hocheng, H., Tai, H.N. and Liu, C.S. (2000), “Assessment of ultrasonic drilling of C/SiC composite material”,
Compos. Part A.-Appl. S, 31(2), 133-142.Harrison, P. and Henry, M. (2006), “Laser processing of polycrystalline diamond, tungsten carbide and a related
composite material”, J. Laser Appl., 18(2), 117-126.Kumar, V. and Khamba, S. (2008), “Statistical analysis of experimental parameters in ultrasonic machining of
tungsten carbide using taguchi approach”, J. Am. Ceram. Soc., 91(1), 92-96.Kumar, V. and Khamba, J.S. (2009), “Parametric optimization of ultrasonic machining of Co-based super alloy
using the taguchi multi-objective approach”, J. Prod. Res. Des., 3(4-5), 417-425.Kumar, V. and Khamba, J.S. (2006), “Experimental Investigation of ultrasonic machining of alumina-based
ceramic composite”, J. Am. Ceram. Soc., 89(8), 2413-2417.Phadke, M.S. (1989), Quality engineering using robust design, Prentice Hall, Englewood Cliffs, New Jersey.Roy, R.K. (1990), A premier on the taguchi method, Van Nostrand Reinhold, New York.Tsao, C.C. (2009), “Grey-Taguchi method to optimize the milling parameters”, Int. J. Adv. Manuf. Tech., 40(1-2),
41-48.Tso, P.L. and Liu, Y.G. (2002), “Study on PCD machining”, Int. J. Mach. Tool. Manuf., 42(3), 331-334Tosun, N. (2006), “Determination of optimum parameters for multi-performance characteristics in drilling by
using grey relational analysis”, Int. J. Adv. Manuf. Tech., 28(5-6), 450-455.Zeng, W.M., Li, Z.C., Pei, Z.J. and Treadwell, C. (2005), “Experimental observation of tool wear in rotary
ultrasonic machining of advanced ceramics”, Int. J. Mach. Tool. Manuf., 45(12-13), 1468-1473.